Which AI crawlers should I allow?

Sort them by what they do with the page rather than by the company that runs them:

• Allow the ones that cite - OAI-SearchBot, PerplexityBot and Claude-SearchBot exist to put a link to you inside an answer.
• Decide separately on the ones that train - GPTBot, ClaudeBot and Applebot-Extended feed model training, which returns nothing to you directly. Google-Extended is the odd one out: it covers Gemini training AND grounding, so blocking it also opts you out of being drawn on inside Gemini answers.
• Leave the user-triggered ones alone - ChatGPT-User, Perplexity-User and Claude-User fetch a page because a person asked for it.
• Never block Googlebot or Bingbot to control AI - AI Overviews and Copilot ride on those indexes, so blocking them costs you ordinary search as well.

Thirteen per cent of Australian small-business sites carry any AI-crawler rule in robots.txt at all^*. Allowing everything is a reasonable answer. It is only a bad one when nobody chose it.